Recognizing User Attentiveness using Deep Learning Methods for Socially Intelligent Agents

Suitable for:

Master project

Description:

Socially intelligent agents and robots are expected to be able to engage in spoken dialogue with human users. An important part of a natural and desirable dialogue is to understand the attention of the interlocutor. This master thesis aims at developing such capabilities for a given dialogue manager framework in our lab. Starting from representations of online language and dialogue processing, you will have to explore and develop methods to dynamically estimate the user’s current level of attention. The project may also include a user study to evaluate the developed approach.

In this project, different deep learning methods can be developed and/or adapted to compare approaches using only speech, eye tracking, facial expressions and gestures in terms of quality and feasibility.

Requirements:

Interest and background knowledge in language and dialogue
Solid programming skills (Python)
Prior knowledge in Deep Learning

Contact:

Hendric Voß